fix(applier): exempt batteries_discharge_mode from the hold-derived 0 W discharge cap

[woopstar]

Summary

Fixes the Huawei Maximum discharging power register oscillating between 0 W and the rated maximum
during a batteries_discharge_mode window while the published recommendation never changes.

Reported against 6.3.0 by @AndersN76 in #978 (comment)
and confirmed by @Extati in #978 (comment).

This PR targets v6.3.0-hotfix so the fix can be released on the 6.3.x line while main moves
toward 7.0.0. The same change is proposed against main separately.

Root cause

applier_caps._primary_battery_hold() derives "the solved plan explicitly holds the primary
battery" from a near-zero (batteries_charged_kwh, batteries_discharged_kwh) pair, and applier.py
turns that into an unconditional 0 W discharge cap. Two facts make that unsound for a schedule
discharge window:

1. planner/soc_simulation.py relabels only force_batteries_discharge / force_export to
   batteries_wait_mode when the simulated discharge is zero — batteries_discharge_mode
   deliberately keeps its label, because it is the user's configured window, not a forced action. So
   a discharge window whose re-solved batteries_discharged_kwh rounds below
   PLANNED_ENERGY_ROUNDING_KWH (0.001 kWh) keeps the discharge label and satisfies the derived
   hold, without the plan ever having decided to hold.
2. batteries_discharge_mode executes as MaximizeSelfConsumption, where this cap is a ceiling
   the firmware ramps within from live house load — not a setpoint. Writing 0 W there disables the
   behaviour the mode exists for and contradicts the Recommendations enum's own contract for it
   ("discharge battery to cover house load").

Because the published recommendation never changed, neither existing damping layer applied:
plan-level hysteresis (#372) stabilises candidate selection, and window hysteresis (#315) only
reacts to a recommendation change (and lets any transition involving a neutral recommendation
through immediately anyway). Nothing guarded this final actuator boundary.

This is not #939/#941: that bug wrote rated max then 0 W inside the same apply cycle. Here the
writes are minutes apart across separate replan/apply events, with the #941 single-write-per-cycle
fix present and working.

Changes

• custom_components/hsem/custom_sensors/applier_caps.py — new _primary_battery_cap_hold(rec):
  _primary_battery_hold(rec) and the recommendation is not batteries_discharge_mode.
• custom_components/hsem/custom_sensors/applier.py — the discharge-cap decision uses the new
  wrapper. _primary_battery_hold() itself is unchanged and keeps its meaning for
  _held_planned_export_is_authoritative() and the batteries_wait_mode working-mode branch (where
  the two are identical, so that branch is untouched).
• docs/planner-spec.md — new "Discharge-mode exemption (issue fix(inverter): batteries_discharge_mode slot writes a 0 W Huawei discharge cap and oscillates against the rated maximum #983)" subsection plus a pointer from
  the discharge-power-cap section.
• docs/troubleshooting-guide.md — new "7i. Maximum discharging power capped at 0 W" with the
  legitimate causes and the capped max discharge power to log pattern.
• .github/memories.md — engineering-log entry, including the two rejected alternatives.
• tests/test_discharge_mode_cap_oscillation.py — new.
• tests/test_coordinator_tracking_{forecast,solar_corrector}.py — unrelated CI back-port, see
  below.

All other 0 W paths keep immediate precedence and are covered by regression tests: EV permission
gating (#797), the planned-EV rate cap, solar-charge-only (#922), the wait-mode reserve floor (#954),
the current_required_battery_kwh SoC guard (#592), and the read-only/degraded gates.
force_batteries_discharge / force_export already bypassed this branch entirely.

Tests

tests/test_discharge_mode_cap_oscillation.py — 16 tests:

• _primary_battery_cap_hold() unit coverage, including the ev_smart_charging relabel and
  wait-mode cases
• no 0 W write on a near-zero-discharge slot; a hardware cap left at 0 W restored to rated max; the
  slot still runs MaximizeSelfConsumption
• an 8-cycle replay of the reporter's timeline with the solved discharge flipping across the
  materiality boundary, asserting exactly one cap write
• precedence regressions for unpermitted EV, permitted-EV rate cap, SoC reserve guard,
  solar-charge-only, and a genuine held Wait slot

Like #939's tests these assert the exact list of writes to the entity, not just the final value —
the pre-fix behaviour passes a final-value-only assertion. Verified failing before the fix and
passing after.

Second commit — stale test call-sites back-ported from main

303eb34 is a separate, test-only commit unrelated to #983, included so CI can actually gate the
6.3.x release.

The v6.3.0 tag shipped three test call-sites that no longer matched
coordinator_tracking.accumulate_forecast_actuals(): one still passed the solar_corrector_processed
argument dropped by #973, and two omitted the required keyword-only update_interval_minutes. Four
tests error out on a clean v6.3.0 checkout and mypy reports three call-arg errors — both
reproduced in a clean worktree of the tag before touching anything, so neither came from this PR.

coordinator_tracking.py is byte-identical between v6.3.0 and main, so this is exactly the
test-only delta already on main (landed there as collateral in #904) — 3 lines, no production
behaviour touched. Happy to split it into its own PR if you'd rather keep this one single-purpose.

Quality gates

Gate	Result
./scripts/quality.sh lint	pass
./scripts/quality.sh typing	mypy 0 errors
./scripts/quality.sh quality	pyright 0 errors, 0 warnings; vulture unchanged
./scripts/quality.sh test	3318 passed, 0 failed
./scripts/quality.sh translations	0 errors (no user-facing strings changed)

Checklist

• docs/planner-spec.md read and updated (semantics of the discharge cap changed)
• Energy balance / SoC bounds / cost identity untouched — this is an execution-layer change only,
  no planner output is mutated
• Read-only / degraded / dry-run gates unaffected (they return before the cap block)
• Tests added covering the changed behaviour and every preserved 0 W path
• Docs updated (planner spec, troubleshooting guide, engineering log)

Fixes #983

🤖 Generated with Claude Code